The Value of Gen-AI Conversations: A bottom-up Framework for AI Value Alignment

TL;DR

This paper introduces a bottom-up framework for aligning conversational AI with human values by analyzing real-world interactions to identify core values and misalignments, offering practical insights for ethical AI deployment.

Contribution

It proposes a novel bottom-up approach using ISO value ontology to analyze conversational logs for ethical value alignment, addressing limitations of top-down methods.

Findings

Identified nine core values in real-world CA interactions.

Detected 32 instances of value misalignment affecting users.

Provided actionable insights for improving ethical AI interactions.

Abstract

Conversational agents (CAs) based on generative artificial intelligence frequently face challenges ensuring ethical interactions that align with human values. Current value alignment efforts largely rely on top-down approaches, such as technical guidelines or legal value principles. However, these methods tend to be disconnected from the specific contexts in which CAs operate, potentially leading to misalignment with users interests. To address this challenge, we propose a novel, bottom-up approach to value alignment, utilizing the value ontology of the ISO Value-Based Engineering standard for ethical IT design. We analyse 593 ethically sensitive system outputs identified from 16,908 conversational logs of a major European employment service CA to identify core values and instances of value misalignment within real-world interactions. The results revealed nine core values and 32 different value misalignments that negatively impacted users. Our findings provide actionable insights for CA providers seeking to address ethical challenges and achieve more context-sensitive value alignment.